Reading Comprehension and Rate: One Column vs. Three Columns

Results

The results of our experiment measured the time to reading completion and the score achieved by subjects on a meta-cognitive test.  Both the time to reading completion and the score achieved by the subject on the test were independently controlled.  Each of the two dependent variables was subjected to an analysis of the raw data.

Individual Results

A bar chart of the read completion times for each of the twenty subjects is shown above.  The one-column completion times are presented in light purple while the three column completion times are presented in burgundy red.  It can be seen that a majority of the subjects read the three column passage faster than the one column passage.  The read times of subjects #1-15 appear to fall into a close range of each other, while subjects #15-20 are not in the same range as #1-15; however, they appear to fall into close range of each other. 

A bar chart of the exam scores for each of the twenty is shown above.  The one-column scores are presented in light purple while the three-column scores are presented in burgundy red.  By observation, it appears that no subject performed overwhelmingly better on either format.  Some performed better on the three column exam, while others performed  better on the one column exam.  Also, the range of score appears to be very broad as some subjects achieved high scores while others achieved low scores.

Subjective Preference Results
The survey of the subject's preference of the one-column or the three-column format reveals a count of nine (9) subjects who favored the one column format and eleven (11) subjects who favored the three column format.  The subjective preference results show that there was no overwhelming preference for either format.

Average, Variance, and Standard Deviation
The time to reading completion and test score data is a follows:

The mean read time and standard deviation are depicted in can be viewed in graphical form by the following illustration:

The results demonstrate the the average read time for three column format is less than for one-column formats.  This result is not consistent with our hypothesis that one column formats would require less time to read.

Test scores were created using a formula that calculated the user's overall "distance" from the set of correct answers.  The formula used to find the distance (RAW_SCORE) are as follows:

The preceding formulas were used to calculate the raw score because each test contained some questions that the subject should respond with 1 and some that the subject should respond with 9.  A total "distance" of 0 from the correct answers means the subject achieved a perfect score on the test, while a total distance of 640 means the subject received the lowest score possible.  The percentage score used for the data analysis is derived from the raw data using the following equation.

((640 - RAW_SCORE)/640)*100%

The preceding equation allows for the test results to be evaluated on a scale of 0-100.  A graphical representation of the average and standard deviation of the Test Score data:

As can be seen in the graph, the average test score and standard deviation for the one-column test and three-column test are almost identical

  
T-Tests
T-tests performed on the data set were used to determine the whether differences of the dependent variables for the given treatments are statistically significant.  Since our experiment only involved a single factor for each treatment, t-tests would suffice in proving whether the difference in read completion time and test scores are statistically significant.  Forming the null hypothesis that one column reading time minus three column reading is equal to zero, and one column test scores minus three column test scores is equal to zero, leads to the equations for the t-value:

 
The result of this computation for the time to read completion variable is a value of t=2.86.  This value is looked up in the t-distribution table and cross referenced with 19 degrees of freedom.   The obtained p-value (p<=.011) shows that the differences in reading times between one and three columns are statistically significant.  

The result of this computation for the test score variable is t=0.10.  Again, cross referencing this value with the t-distribution table at 19 degrees of freedom yield a p-value of (p<=.246).  This p-value shows that the differences in average test scores between one and three columns are not statistically significant.   This validates the observation of the bar chart of "Average Test Scores" as the two bars are almost identical.

Read Time vs. Test Scores

Plotting the read time against the test scores reveals a slightly positive trend between test score and read completion time.  This indicates that subjects who spent more time reading the passages comprehended more, and thus, performed better on the test.   There were only a few aberrations of fast reading and high score, but these could be accounted to chance since each test only consisted of five questions.

Department of Computer Science: Direct questions and comments to the student editorial team